Operational range estimation for deploying mobile robots
If you have ever driven a car, more specifically an electric vehicle (EV), you would have considered the mileage when rolling out of your drive way. This is an essential consideration to avoid getting stranded owing to complete battery depletion. Similar concept applies to most modern cars powered by gasoline or diesel where the dashboard shows estimated mileage.
Have you ever wondered if there was a possibility to estimate the mileage of your mobile robots too? After all, they too have rechargeable batteries and if the battery runs out, a drone might just fall out of the sky which isn’t desirable.
Determining the operational range of a mobile robot is one of the most important aspects for planning, calibrating, and deploying it in real world.
In this blog post, we will go over the basics of operational range estimation for mobile robots. We’ll discuss the importance of figuring out the operational range, some of the methods used to estimate it and give an overall understanding of how to approach the process in your own projects and applications.
What is operational range estimation?
Operational range estimation is a process of systematically determining the operational range of a mobile robot where the range is a measure of distance that a robot can cover on a single discharge cycle.
The operational range of a robot is typically affected by various factors:
Battery: how strong the battery is, how much charge can it store and how efficiently does it deplete over the single discharge cycle.
Agility: the more agile maneuvers a robot needs to execute, the higher would be the energy consumed thereby reducing the range- similar to frequent speeding/breaking that reduces the mileage of a car.
Payload: the more sensors that a robot has, the more energy they will consume leaving lesser energy from the battery to be able to move around. This is similar to reduction in mileage with increase in number of passengers and luggage in case of cars.
Speed: There is usually an optimal speed (nominal velocity) which consumes the minimal amount of energy possible. Any speed higher than this nominal value would increase energy consumption and reduce range. On the other hand, any speed lower, may reduce the chances of the robot completing the task on time.
Disturbances: There might be unforeseen disturbances in the operational environment such as sudden appearance of pedestrians, wind gusts, terrain changes, elevation changes etc., which might spike the energy consumption.
To better understand the concept of operational range, let’s start with a simple example. Imagine that you have a robot and you want it to navigate an office hallway. To do this, it needs to have the ability to sense its environment, understand its surroundings, and react accordingly.
However, that’s not all. The robot also needs to know how far it can go in a given direction and for how long, as well as its optimal speed to reach its destination. This is where operational range estimation comes into play.
While endurance is a concept popular for drones and defines how long can a drone fly before returning to base, the operational range estimation is generalized to all sorts of robots.
Why consider operational range estimation for deployment?
Estimating the operational range of a robot can be beneficial for a variety of applications and scenarios. Knowing the operational range of a robot will help ensure that it can reach its destination and make sure it has the ability to return to base.
One of the most common benefits is that it can be used to improve situational awareness, allowing the robot to anticipate potential obstacles and make smart decisions in order to reliably reach its destination.
It can also be useful in optimizing the paths of the robot, helping it to avoid maneuvers that would increase the energy consumption and find the energy-efficient route to its destination.
Ignoring the operational range would likely result in total battery depletion and complete immobilization in the middle of the mission, which isn’t desirable.
Methods for estimating operational range
There are prominently two approaches for operational range estimation for mobile robots:
Offline one-shot range estimation: Think of this as doing a rough estimation of the range, given a robot, battery characteristics, and some prior information about the mission. All else being the same, this is the best the robot can do but of course, when the robot actually deploys, things to remain exactly as you imagine them to be. This is the limitation of an offline one-shot model that does not update as per situation.
Online recursive range estimation: To overcome the limitation of the offline one-shot learning, there is a need to develop a recursive estimator that not only accounts for how the energy is being consumed as the mission unfolds, but also estimates the future demands to inform the robot of the probability to return to base if it continues to do what it is doing.
I have delved into both of these models and you can read my publications to get detailed technical insights in the current state of art in operational range estimation.
Key components for operational range estimation
Accurate operational range estimation requires careful analysis of the robot’s behavior in the environment. The following components should be taken into consideration:
Battery depletion model: This model defines how the energy from the battery is depleted over time. This can be a heuristic model or can be a complex non-linear model accounting for the chemical dynamics of the battery itself.
Ancillary energy model: Ancillary components such as sensors and on-board processors all consume energy from the same power source- the on-board battery. Thus, model systematically identifies how each ancillary component consumes energy over time. For sensors, this is heavily dependent on the operational frequency.
Maneuvering energy model: The battery also supplies energy to the motors that are then used to propel the robot forward. This is the most crucial aspect for navigation because sensors help the robot perceive the environment and consume the energy in the process yet do not help cover the distance directly.
Energy loss model: As a part of the system identification, not only is the energy consumed by various components modeled, even the energy lost owing to friction, resistance, etc., needs to be accounted for. The overall system efficiency informs how well the robot is utilizing the energy provided by the battery.
Robot’s mechanical model: Operational range estimation is a very mechatronics related problem- there is component-wise energy consideration and also mechanical constraints stemming from the dynamics of the robot platform itself.
All these models above need to work in tandem to form a composite operational range estimation model which could either work offline or online as per the situation.
Accurately estimating a mobile robot’s operational range is essential for understanding its capabilities and performance. It helps to ensure the robot will reach its desired destination, accomplish its task and return to base while avoiding complete immobilization, except from sudden hardware failures.
Over the years, I have been working consistently on developing various offline and online range estimation models that generalize over various robot platforms- commercial and custom built, alike. There is still room for improvement along these lines if we were to obtain robust robot capable of safe real-world deployment.